This paper presents the development of optimal machine learning-based models for the prediction of the compressive strength (CS) of self-compacting concrete (SCC). The models included regression models that are multi-parametric and optimal artificial neural networks (ANN). Gradient descent search was adapted in determining optimal regression models while Akaike’s information criteria (AIC) blended with the best training strategy were used in the explorative search for optimal ANN topology. Around 900 data instances of SCC mix were used to find the models. The inputs used for this are quantities of mix ingredients such as quantities of cement, water–powder ratio, fine aggregate, fly ash, and coarse aggregate. Twenty-eight day’s compressive strength formed the output. Results of the evaluation of the models indicated that, for the data considered, optimal ANN model of topology 5-13-1 is found to be the best with a high R2 value at 0.9 while the optimal regression models showed moderate prediction accuracy with R2 in the range of 0.7–0.75.